On 04.06.2008 16:53, ron minnich wrote:
So, I went to apply and test this patch, but I guess, on reflection, I'm pretty uncomfortable with it. Why? Because of this:
- hide_vpci(0x800079C4);
This is even worse than putting stuff like this in the code dev = dev_find_pci_device(PCI_VENDOR_ID_AMD, PCI_DEVICE_ID_AMD_CS5536_UDC, 0); if (dev) pci_write_config32(dev, 0x7C, 0xDEADBEEF);
because at least the latter form doesn't hard code pcidevfn.
I think you are right. hide_vpci should take a dev pointer. So let's try to rework with that change to hide_vpci and we can then bring this patch in.
hide_vpci then becomes pretty simple in fact ...
Please note that my patch does only change those occurences where
pci_write_config32(dev, 0x7C, 0xDEADBEEF);
was unusable. The rest of the changes is just better commenting, avoidance of variable reuse and so on.
My original patch had the pci_write_config32 -> hide_vpci transformation (bad), the updated patch (dated 08.05.2008 02:12) didn't, that way I can send a followup patch which introduces hide_vpci_dev and cleans up the rest once you ack this one.
Regards, Carl-Daniel